Method for detection of defects lacking distinct edges

Abstract

A method of image processing for visually inspecting a workpiece. The method compares the brightness (15) at each location within an image of the workpiece to the equivalent location within an image of an idealized workpiece. The inspection depends only on local brightness differences between the two images. The method can detect defects which have no distinct edges. Finally this method can detect defects which are small enough that the resolution of the image will show these small defects only as a single point of light (12) or dark (27).

I claim: 
     
       1. A method for detecting defects which lack distinct edges, comprising: providing a graylevel representation of a taught image of a defect free workpiece;   providing a graylevel representation of a run image of a workpiece being inspected, the run image revealing a defect which lacks distinct edges;   registering the run image by mechanical and optical means to scale, rotation and spatial position of the taught image;   refining the registration between the run image and the taught image by relaxation of the graylevel representation of at least one of the two images;   providing a means to determine a mean brightness level of the run image;   mapping values of the graylevel of each pixel of one image based on a comparison of the mean brightness level of the two images; and   comparing the graylevel representations of the taught image with that of the run image to create a representation of a composite image which highlights the defect which lacks distinct edges.   
     
     
       2. The method for detecting defects which lack distinct edges of claim 1 further comprising: mapping the graylevel representation of the run image based upon the relative location of a local minima in a graylevel histogram which serves to characterize the taught image when compared to the location of an equivalent local minima in the graylevel histogram which serves to characterize the run image.   
     
     
       3. The method for detecting defects which lack distinct edges of claim 1 further comprising: mapping of the graylevel values using a linear scaling.   
     
     
       4. A method for detecting defects which lack distinct edges, comprising: providing a graylevel representation of a taught image of a defect free workpiece;   providing a graylevel representation of a run image of a workpiece being inspected;   registering the run image by mechanical and optical means to scale, rotation and spatial position of the taught image;   refining the registration between the run image and the taught image by relaxation of the graylevel representation of at least one of the two images;   providing a means to determine a mean brightness level of the run image;   mapping values of the graylevel of each pixel of one image based on a comparison of the means brightness level of the two images;   comparing the graylevel representations of the taught image with that of the run image to create a representation of a composite image which highlights defects including defects which lack distinct edges in the run image; and   identifying defects by recording locations within the composite image which fall outside a predetermined range of acceptable graylevels.   
     
     
       5. The method for detecting defects which lack distinct edges of claim 4 further comprising: relaxing the graylevel of the run image with a combination of graylevel morphological erosion and graylevel morphological dilation.   
     
     
       6. The method for detecting defects which lack distinct edges of claim 4 further comprising: refining the registration between the run images and the taught image by relaxation of local brightness maxima and minima within the graylevel representation of the run image alone.   
     
     
       7. The method for detecting defects which lack distinct edges of claim 4 further comprising: refining the registration between the run image and the taught image by relaxation of local brightness maxima and minima within the graylevel representation of the taught image alone.   
     
     
       8. The method for detecting defects which lack distinct edges of claim 4 further comprising: eroding the representation of the composite image using a feature erode of defects having a brightness level which is greater than a predetermined range of acceptable graylevels so as to eliminate bright defects which are smaller than a predetermined size.   
     
     
       9. The method for detecting defects which lack distinct edges of claim 4 further comprising: dilating the representation of the composite image using a feature dilate of defects having a brightness level which si lower than a predetermined range of acceptable graylevels so as to eliminate dark defects which are smaller than a predetermined size.   
     
     
       10. A method for detecting defects which lack distinct edges, comprising: providing an image of a defect free workpiece;   providing an image of a workpiece being inspected having a defect which lacks distinct edges;   adjusting values of a graylevel representation of the workpiece being inspected based on the relative location of a local minima in a graylevel histogram which serves to characterize the defect free workpiece image and an equivalent local minima found on the graylevel histogram which serves to characterize the image of the workpiece being inspected;   aligning the representation of the image of the workpiece being inspected to the defect free image by mechanical and optical means;   refining the alignment between the representation of the image of the workpiece being inspected and the representation of the image of the defect free workpiece by a relaxation of the graylevel representation of at lest one of the representations of the images;   comparing the graylevel representations of the images of the workpiece being inspected with the representation of the image of the defect free workpiece to create a representation of a composite image which highlights the defect which lacks distinct edges in the run image; and   identifying defects including the defect which lacks distinct edges by recording locations within the composite image which are outside a predetermined range of acceptable graylevels.   
     
     
       11. A method for detecting defects which lack distinct edges, comprising: providing a representation of an image of a defect free workpiece;   providing a representation of an image of a workpiece being inspected, the workpiece including a defect which lacks distinct edges;   registering the representation of the image of the workpiece being inspected by mechanical and optical means to scale, rotation and spatial position of the representation of the image of the defect free workpiece;   refining registration between the representation of the image of the workpiece being inspected and the representation of the image of the defect free workpiece by relaxation of a graylevel representation of the image of the defect free workpiece;   scaling values of the graylevel representation using a linear scale factor of the representation of the image of the workpiece being inspected based on the relative location of a local minima in graylevel histograms which characterize the representation of the image of the defect free workpiece and the representation of the image of the workpiece being inspected;   comparing the graylevel representations of the image of the defect free workpiece with that of the image of the workpiece being inspected thereby creating a representation of a composite image which highlights defects in the run image;   eroding the representation of the composite image to eliminate defects which are smaller than a predetermined size; and   identifying defects, including the defect which lacks distinct edges by recording locations within the composite image which are outside a predetermined range of acceptable graylevels for the composite image.